Pasting into a DataTable overwrites `data_previous` with new data

Environment

• Python 3.10.7

• pip list | grep dash

  dash                 2.7.0
  dash-core-components 2.0.0
  dash-html-components 2.0.0
  dash-table           5.0.0
  

• same on all browsers I tested:

  • on MacOS 12.6:
    • Chrome 107.0.5304.110
    • Version 16.0 (17614.1.25.9.10, 17614)
    • Firefox 105.0.3 (64-bit)
  • on Windows 10 Pro
    • Microsoft Edge 107.0.1418.52 (Official build) (64-bit)

Describe the bug

When doing a paste action (as in copy&paste), two things happen:

• data property gets updated
  • ☝️ this is as expected
• data_previous property takes on the pasted value
  • ☝️ this seems to be a bug

Expected behavior

After a paste, I expect data_previous property to contain last the data before the paste.

Why is this Important

Normally, comparing data with data_previous makes it is easy to see what has changed. This is important when values in different columns are inter-dependent. This bug makes it impossible (without non-pretty workarounds) to tell which values were pasted into the table.

Screen Recording

https://user-images.githubusercontent.com/1173748/202705073-96d56e50-b59f-4bab-8a77-df9e9bd1f352.mp4

Example App

import dash
import pandas as pd
from dash import dash_table, dcc, html, Output, Input, State

df = pd.read_csv('https://raw.githubusercontent.com/plotly/datasets/master/solar.csv')

app = dash.Dash(__name__)

app.layout = html.Div([
    dash_table.DataTable(
        id='table',
        columns=[{"name": i, "id": i, "editable": True} for i in df.columns],
        data=df.to_dict('records'),
        editable=True,
    ),
    # counter for callback calls, to make sure the callback is not running multiple times:
    html.H4("Callback count:"),
    dcc.Input(id="count", value=0),
    # current and old data for inspection:
    html.H4("Current data:"),
    html.Pre(id="data", children=""),
    html.H4("Previous data:"),
    html.Pre(id="prev_data", children=""),
])


@app.callback(
    Output("data", "children"),
    Output("prev_data", "children"),
    Output("count", "value"),
    Input("table", "data"),
    State("table", "data_previous"),
    State("count", "value"),
)
def show_data_and_prev_data(data, prev_data, count):
    return str(pd.DataFrame(data)), str(pd.DataFrame(prev_data)), int(count) + 1


if __name__ == '__main__':
    app.run_server(debug=True)

Run the provided Python example app and reproduce the paste behavior while comparing the DataTable data and data_previous properties. Trace the DataTable update path that handles pasted edits; done means data contains the pasted values while data_previous still contains the complete data from before the paste.